Failed to Transform Image

Trial after trial, I cannot get the remove module to work. I’m trying to remove a drone shadow from the ground. Pretty simple. I create a foreground matte of the shadow - track it, and a background matte of the ground covering it, track that. I create a clean plate in Remove, edit it in photoshop, save and head back into mocha, click render frame… and get the error. Every time. I’ve deleted and retracked, recreated the plate several times, from different frames, all to no avail. Tried DPX, TIFFS, even JPEGS. They’re always solid tracks. I’ve also tried remove on completely different clips with the same results. There must be a setting wrong somewhere.

Error message reads: “Failed to transform image /Volumes/…etc.”

Using mocha pro 4.1.3 on OSX 10.8.5/Mac Pro 2 x 2.4 GHz Quad-Core/ATI Radeon HD 5770 1024 MB


Thank you,





Hi David,

Can you tell me what the Bit Depth and Format settings are for your clip in the Clip tab? You can find these parameters on the bottom left of the page.


Hi Martin,

Clips settings: 23.976fps DPX sequence at 2048x1080, a bit depth of 96, RGB, Float 32. They are in Log format, with convert to float checked.

Not sure if it matters, but when I read the files, they come in at .9375 PAR, which is incorrect. They should be square pixels. I didn’t prep the DPXs, but the post house that sent them to me claim they should be square pixels as well. When I read them in Invisor, they also claim to be square. Could mocha’s misread of the PAR metadata be part of the issue?


Thanks for any help.


After some more experimentation, it looks like the error is in selecting “convert to float”. Unchecked, the remove module works beautifully. Given I am compositing linearly in AE in 32 bit float (for a feature) I thought this was a similar setting. Admittedly I don’t have a complete grasp of 32 bit float processing, other than it improves the performance of lighting and compositing operations, especially in linear mode.

What is the detailed purpose of this feature in mocha? It’s not well documented other than “…increasing the precision of compositing operations.” If I can do the remove composite using a linear illumination module without ‘convert to float’ checked, what purpose does it serve?

Obviously my concern is in not losing information. I have to output back to 10bit DPX files from AE.

Thanks for any insight.